Xinxiu Tao - Reliability of Changes in Cloud Environment at PaaS level

Équipes : 


Jury :

  • Françoise Baude, professeur, CNRS (rapporteur)
  • Lionel Seinturier, professeur, Inria (rapporteur)
  • Daniel Hagimont, professeur, IRIT/ENSEEIHT (examinateur)
  • Thomas Ledoux, enseignant-chercheur, IMT Atlantique (examinateur)
  • Fabienne Boyer, maître de conférences, LIG/UGA (directrice de thèse)
  • Noël de Palma, professeur, LIG/UGA (co-encadrant)
  • Xavier Etchevers, ingénieur de recherche, Orange Labs (co-encadrant)


Microservice architectures are considered with much promises to achieve DevOps in IT organizations,  because they split applications into services that can be updated independently from each others.  But to protect Service Level Agreement (SLA) properties when updating microservices, DevOps  teams have to deal with complex and error-prone scripts made up with many management operations. In this thesis, we propose an update framework that leverages an architecture-based approach to  provide an easy and safe way to update microservices. Instead of scripts processing management operations,  update processes are defined as sequences of generic elementary changes being applied on an architectural  model of a microservice application. Simply put, this architectural model reflects how microservices  are deployed on Cloud sites and how they are configured, following a model@runtime abstraction.  Thanks to this architecture-based approach, DevOps teams just have to specify a target architecture for  their application to update, they no more need to determine the precise pipeline of commands to process.  The proposed framework has been evaluated in real conditions, through updating clones of microservice-based IT applications  in live production at Orange.